Heart Disease Classification

Table of Contents

1. Description
2. Dataset
3. Installation
4. Usage
5. Project Structure
6. Contributing
7. License
8. Analysis
9. Heart Disease Prediction Dashboard.

Description

This project aims to classify the presence of heart disease using the Heart Disease UCI dataset. It explores various machine learning algorithms to predict whether a patient is likely to have heart disease based on medical attributes such as age, cholesterol levels, blood pressure, and more.

The goal is to demonstrate expertise in:

• Data preprocessing and exploratory data analysis (EDA).
• Building and evaluating classification models.
• Developing an interactive dashboard for visualizing predictions.

Dataset

The dataset is sourced from Kaggle: Heart Disease UCI Dataset. It contains 14 features, including both numerical and categorical variables, and a target column (target) indicating the presence (1) or absence (0) of heart disease.

Installation

To run the project locally, follow these steps:

1. Clone this repository:

   git clone https://github.com/oscarTMa/heart-disease-classification.git
   

2. Navigate to the project directory:

   cd heart-disease-classification
   

3. Install the required Python libraries:

   pip install -r requirements.txt
   

Usage

Notebook:
The initial analysis and modeling will be conducted in a Jupyter Notebook, hosted in Google Colab. The notebook file can be found in the notebooks/ directory.

Dashboard: An interactive dashboard will be built using Streamlit to showcase predictions and insights. Instructions for running the dashboard will be added in later stages.

Project Structure

heart-disease-classification/
│
├── README.md # Project documentation
├── .gitignore # Ignored files and directories
├── requirements.txt # Python dependencies
├── notebooks/ # Jupyter/Colab notebooks for analysis
├── src/ # Source code for preprocessing and models
└── app.py # Streamlit app for the interactive dashboard

Contributing

Contributions are welcome! Feel free to open issues or submit pull requests to improve this project.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Analysis (see the notebook)

1. Logistic Regression

• Accuracy: 0.8533

• ROC AUC: 0.8556

• Classification Report:

  • Precision: The model performs well in both classes, with a precision of 0.80 for class 0 and 0.90 for class 1. This indicates good performance in classifying both classes.
  • Recall: The recall for class 0 is high (0.87), meaning the model is correctly identifying many of the negative cases. However, the recall for class 1 is slightly lower (0.84), indicating that some positive cases are being missed.
  • F1-Score: The F1-score for class 1 is higher (0.87) compared to class 0 (0.83), suggesting the model performs slightly better at predicting class 1.

• Confusion Matrix: The model has some false positives and false negatives, but overall, it is fairly balanced.

2. Decision Tree

• Accuracy: 0.8370
• ROC AUC: 0.8325
• Classification Report:
  • Precision: Similar to Logistic Regression, the model shows good precision for both classes, with a slight difference (0.81 for class 0 and 0.86 for class 1).
  • Recall: The recall is fairly balanced, with 0.81 for class 0 and 0.86 for class 1. However, recall is not as high as Logistic Regression's for class 1.
  • F1-Score: The F1-score is quite similar for both classes (0.81 for class 0 and 0.86 for class 1), indicating a reasonable but not exceptional performance.
• Confusion Matrix: The model shows some errors (false positives and false negatives), but it is performing reasonably well.

3. Random Forest

• Accuracy: 0.8750
• ROC AUC: 0.8725
• Classification Report:
  • Precision: The model shows excellent precision for class 1 (0.90) and good precision for class 0 (0.85).
  • Recall: The recall for class 1 is very high (0.89), and recall for class 0 is 0.86, indicating a good balance in capturing both types of classes.
  • F1-Score: The F1-score is also high for both classes, with 0.85 for class 0 and 0.89 for class 1. The model demonstrates balanced performance across both classes.
• Confusion Matrix: The confusion matrix shows that the model makes few errors, achieving solid performance with minimal false positives and false negatives.

4. XGBoost

• Accuracy: 0.8587
• ROC AUC: 0.8603
• Classification Report:
  • Precision: Precision is quite high for class 1 (0.90) and decent for class 0 (0.81). The model performs well in classifying the positive class.
  • Recall: The recall for class 0 is 0.87, and for class 1, it is slightly lower (0.85). This suggests that the model is good at capturing negative cases but may miss some positive cases.
  • F1-Score: The F1-score is 0.88 for class 1, which is the best among all models. For class 0, it is 0.84, which is also very good.
• Confusion Matrix: The confusion matrix shows solid performance, but the model has some errors in predicting the positive class (false negatives).

Conclusions and Recommendations:

• Best Model: The Random Forest model appears to perform the best overall with the highest accuracy (0.875) and ROC AUC (0.8725). It also maintains a good balance between precision and recall for both classes.
• Good Model: XGBoost also provides strong results, with high precision and recall for class 1, though it is slightly inferior to Random Forest in terms of accuracy.
• Simple Models: Logistic Regression is fast and effective but has a slightly lower recall for class 1 compared to other models. It also has lower accuracy compared to Random Forest and XGBoost.
• Decision Tree: While the Decision Tree model has a good accuracy, its performance is slightly lower than the other models, particularly in terms of ROC AUC and balance between precision and recall.

To improve performance, Random Forest or XGBoost are the most robust models, which we can explore further, possibly with hyperparameter tuning to achieve better results.

Heart Disease Prediction Dashboard

Explore the interactive dashboard here.